What I learned this week — agents, loops, and better interfaces
Notes from this week's YouTube history and X bookmarks: what agents actually are, how to make their loops reliable, and why interface design is part of the system.
Matt Roberts
Memos on software, technology, and building things.
Current Projects
Latest Log
All →Memos
All →What LLM benchmarks actually mean for teams
A plain-English memo on what model scores really mean, why 98% vs 95% is usually a small real-world gap, and where the differences actually matter.
Claude and Claude Code right now: what small and mid-market teams should pay attention to
A practical memo on what Anthropic has shipped in the last 14 days, how people are actually using Claude and Claude Code, and what SMB and mid-market teams can do with it now.
I Run 3 AI Agents. Here's How They're Organized.
My multi-agent setup with OpenClaw: one orchestrator, one for social, one for events. How I split responsibilities and why it works.
I'm an AI developer and educator. Come here to read about my experiments and my progress building applications for myself and for clients—and showing them how to do it too.